Removing High-Frequency Short Portfolios to Improve CSI 300 Enhancement

Summary

This report examines whether removing stocks selected by high-frequency multi-factor short portfolios can improve a CSI 300 index-enhancement strategy. It compares three approaches: combine standardized factor scores, use a regression model to forecast returns and exclude the lowest-ranked stocks, or build individual factor short portfolios and combine them. It recommends orthogonalizing high-frequency factors against style and industry effects, and forming the composite outside the index constituents to capture effects among other stocks.

The reported backtests say all three methods improved excess returns under selected settings. At a 5% short threshold, the standardized-score method with ICIR weighting raised annualized excess return from 15% to 16.7%; regression reached 16.3%, and portfolio combination reached 16.0%. These are report-specific historical results, not evidence of future performance. Regression was sensitive to factor choice and required screening for correlated inputs; portfolio combination also required tuning both the short threshold and the number of component portfolios.

Key ideas

• The report compares score aggregation, return regression, and combined single-factor short portfolios.
• Orthogonalizing factors against style and industry effects is recommended for more stable short effects.
• The score method combines standardized factor values and removes the lowest-ranked stocks.
• Regression requires factor screening because results are sensitive to factor selection and correlation.
• The historical tests report higher CSI 300 excess returns under selected configurations.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.